DRAGON

DRAGON is an AI-augmented SSH and serial terminal with a local inference, redaction, and audit layer in the session path.

Install DRAGON, start the trial, open SSH and serial sessions, import existing sessions, point inference at a model, and ask the copilot.

Point DRAGON's copilot at a local or cloud model — LM Studio, Ollama, OpenAI (ChatGPT), Claude, or any OpenAI-compatible endpoint — with the exact settings for each.

DRAGON's three-tier topology, the loopback WebSocket contract and per-launch token, the hot-path tee, and the in-process event bus.

The transport broker for SSH and serial, full SSH authentication and ~/.ssh/config resolution, TOFU host keys, and the capture engine.

The saved-session tree, drag-and-drop organization, session properties, quick search, the Vault credential manager, and PuTTY and SecureCRT import.

The xterm.js terminal surface, grid layouts with per-tile PTY and scrollback, focus behavior, and built-in keyword highlighting.

The inference orchestrator, embedded and endpoint model modes, inline ask, ambient analysis, auto-analyze, command classification, and hybrid retrieval.

The two-stage secret redaction pipeline, the append-only hash-chained audit log, the transparency pane, and configurable audit storage.

Daemon launch flags and environment variables, the Settings page, persisted settings and hot-swap semantics, audit storage, and the data directory.

DRAGON's offline-first security posture — no telemetry, two-stage redaction, auditable AI with provenance records, keychain credentials, and daemon hardening.

The 14-day trial, offline Ed25519 license activation, what the license gates, signed installers, the two distribution modes, and the update channel.